Business Week's Aug. 26 issue featured a story, "Has the Net Finally Reached the Wall?", on the traffic jams resulting from Net overload. The Aug. 25 guest in this event was Peter Coy, technology editor of Business Week and author of that story.
OnlineHost: Copyright 1996 America Online, Inc.
OnlineHost: Material entered into AOL by persons other than those identified as Business Week's employees or authorized representatives, acting on behalf of Business Week, is material for which Business Week assumes no responsibility.
OnlineHost: Welcome to Business Week Online and the first of back-to-back conferences tonight. This one is on Internet traffic jams, those slowdowns and even crashes (remember Aug. 7 on AOL?) that plague your online life. The guest is Peter Coy (PeterCoyBW), technology editor of BW and author of an Aug. 26 BW story on the Net's growing pains. The moderator is Bob Arnold (BobABW), editor of BW Online and senior editor of the magazine.
BobABW: Hi, Peter, and welcome!
PeterCoyBW: Hi, Bob. Glad to be here.
BobABW: And welcome to you in the audience. Thanks for attending on this very nice evening in the East. Peter, could you start us off by telling us a little about what your recent story said on the subject of traffic jams on the Internet, specifically the Web?
PeterCoyBW: Sure, Bob. I think this topic tonight is something a lot of AOL subscribers can relate to. I'd be interested in hearing what some of tonight's participants experienced on Aug. 7 when the AOL service was down for 19 hours. I think it may have been a sign of things to come for the Net as a whole. See, the problems weren't really unique to AOL. In fact, they involved the type of gear, such as routers, that the entire Internet uses.
BobABW: Peter, here's the first audience question, from RuffnBuff.
Question: Won't any bottleneck on the Internet be eventually alleviated by fiber optics?
PeterCoyBW: No, it's not that simple -- because bandwidth is only one problem. There's also, for instance, the problem of router capacity. And server capacity. That's just to name two things. In fact, if you fill up the pipelines more, you'll notice the problems at the end points even more.
BobABW: Up next, with another bandwidth question, is JackB:
Question: Will better communications -- more bandwidth and/or faster connections -- help or hurt the Net? Will it be like the new freeway that creates new traffic jams almost as soon as it's opened?
PeterCoyBW: I think that's a very perceptive point. First, more bandwidth is definitely a positive. But it's not a cure-all.
BobABW: Here's one of those AOL questions you solicited, Peter -- from Cuckoo95.
Question: Won't the problems always be worse on AOL, since all 6 million members pass through a single chokepoint at the machines in Virginia?
PeterCoyBW: I'd be interested if someone from AOL is listening in to comment on that. I don't claim to be an expert on the AOL architecture. There's nothing inherently wrong with locating lots of server capacity in one place, as long as you have redundant capacity there, and redundant connections to it.
BobABW: Peter, another AOL followup question.
Question: AOL plans to spend $300 million marketing the service in the next two months. They've spent more than $200 for every member they've added -- yet just over $100 million upgrading their equipment. Isn't that where the problem is?
PeterCoyBW: I feel bad here talking so much about AOL, since they're our hosts for this session. And again, I don't know that much about their particular problems. I was only using them in the article as a symbol of the difficulties of the Internet as a whole.
BobABW: A reminder: AOL's Steve Case held a conference a few days back [Aug. 11] on the AOL blackout. The transcript of that is in AOL Live. Now, back to the matter at hand. This one's a little obscure -- do you get it, Peter?
Question: Is the problem stemming from ISPs or NAPs?
PeterCoyBW: I can take a stab at that. It seems to me that the most frequent complaints concern the Internet Service Providers, whereas the NAPs [Network Access Points] -- which provide the connections between the ISPs -- are more invisible to the general public. But it's true that the NAPs are a big source of congestion. That's one reason that lots of ISPs, the big ones mainly, are forming links between themselves that bypass the NAPs. I'd say that's the wave of the future.
BobABW: Your answer leads to this next followup, from RuffnBuff.
Question: Would it be better for users to use smaller ISPs?
PeterCoyBW: I can't see why smaller would be better. Some of the little guys claim more personalized service. But in general, the trend is for concentration into fewer, bigger ISPs.
BobABW: Matt01778 has what I would characterize as a pessimistic question.
Question: Will this mean that a 28.8 modem will work as good as a 14.4 once these traffic jams start?
PeterCoyBW: Yeah, that's pessimistic, all right. I think the traffic jams have begun. And there are times when you get 14.4-like performance or worse out of a 28.8 modem, or even out of a 56-kilobit link. That's speaking from personal experience.
BobABW: Density 21 is next, with a very good question.
Question: What will happen when the cable companies enter the Internet market with the greater access speeds?
PeterCoyBW: Good question. It will certainly give people bigger entry points. But it will add to the problem of congestion, because people will place more demands on the network than ever.
BobABW: Peter, here's an observation from Cuckoo95. Any reaction?
Comment: The router problem was unique to AOL in that a router problem affected all 6 million users, nationwide -- the nation's largest Internet provider has about 400,000 members.
PeterCoyBW: True enough. I just meant that router problems in general aren't unique to AOL.
BobABW: Sir Good is next, Peter, with this one.
Question: What's the traffic jam in respect to -- Internet bandwidth or number of users a particular system can handle?
PeterCoyBW: Both. Either can be a source of congestion -- local or backbone. Glad to see that this topic is attracting lots of interest.
BobABW: Peter, as a former telecom editor, you're the one to answer IrvLustig.
Question: MCI and Sprint are two major backbones. Doesn't the Internet phone application threaten their core business (long distance)?
PeterCoyBW: That is a tough one. It certainly does change their business. It means that the price they can collect from a long-distance call is headed down -- way down. But on the other hand, it should lower the cost of carrying long-distance calls by a similar amount. So they might be able to come out O.K. Stay tuned on that one.
BobABW: Looks like AdamB586 tuned in late. But he has a good question. The only breakdown we've mentioned tonight is AOL's. What other signs are there of stress on the Net? And here's Adam.
Question: What does it mean -- traffic jams?
PeterCoyBW: I'd say that traffic jams include servers not responding. It also includes servers that do respond, but it still takes forever to, say, have a page come up on your screen. That's why people call the WWW "slow television" or the World Wide Wait.
BobABW: Peter, what's the hard evidence (aside from anecdotal) of a Net slowdown? Any?
PeterCoyBW: I spent many hours trying to come up with good statistics on this. One good place, if you're interested, is the Web site of MERIT, which oversees -- or rather coordinates -- the North American Network Operators Group. Sorry I don't have their URL right now. They keep lots of statistics. But unfortunately even that is limited because the various operators are reluctant to share all their performance data with each other. That's one thing that will have to change for the Net to really be a mature medium of communication.
BobABW: And now comes JPascy, referring to 14.4 modems.
Question: Is a 14.4 modem soon to become outdated?
PeterCoyBW: Yes.
BobABW: JDenyer46 is up next. And let me say that AOL 3.0 responds faster than 2.7, so you might want to go to keyword Upgrade and download it.
Question: It seems to take forever to access many databases and files on the Internet, and sometimes I can't at all. Is this a problem with my system or the result of a traffic jam? I'm using version 2.7 of AOL with the improved Web browser.
PeterCoyBW: Well, BobABW just gave you part of your answer. I'd say that, yes, the other part is probably system slowdown.
BobABW: CInga9223 has a very practical question.
Question: When is the best time to go online and miss these traffic jams?
PeterCoyBW: I think that depends on what you want to do. Different Web sites are popular at different times. As for overall backbone congestion, I think daytimes are the most crowded, when both the East and West Coasts are going full steam.
BobABW: IrvLustig is here, with a follow-up question that suggests a (gulp) conspiracy.
Question: Isn't it in MCI's and Sprint's interest to cause the traffic jam?
PeterCoyBW: Frankly, I don't see why.
BobABW: Cuckoo95 is back, with a tough question, Peter.
Question: AOL isn't really a symbol of the Internet's capacity difficulties, since their architecture (and problems) are unique to AOL. It's apples and oranges.
PeterCoyBW: Well, that could be the topic of a long and very involved debate. If this is back to the point that all the servers are in one place in the country, I still fail to see why that should be an issue, given sufficient redundancy in server capacity and connections to it. I know AOL has begun turning its attention to reliability more and more. We shall see.
BobABW: Here's an observation from DaveToppe. Reaction, Peter? (IRC, I assume, is Internet Relay Chat).
Question: I have been using IRC for quite some time, and it looks like a power outage when everybody gets split.
PeterCoyBW: Sorry, I don't use IRC, so I can't say.
BobABW: Bary0011 has a question to which we'd all like to know the answer.
Question: What's to stop other crashes like the one on Aug. 7?
PeterCoyBW: One of the people we quoted in our article was a professor from Columbia University who made the point that whenever you have a system that's undergoing big changes while it is in operation, you run the risk of a breakdown. It's like trying to change the jet engines on a 747 while it's halfway across the Pacific Ocean. So it would be very difficult to promise no further such problems.
BobABW: These AOL questions just won't disappear.
Question: I've heard using the Web via AOL is slower than doing it through an ISP. Isn't that "traffic jam" actually caused by the service's hardware, and not the number of users?
PeterCoyBW: Maybe. I don't know about that.
BobABW: One second, folks. Peter wants to raise an issue no one has yet asked about. O.K., Peter, shoot!
PeterCoyBW: One issue that the article got into that so far no one has raised is searching the Web. As many of you well know, there are lots of amazing Web search engines that spit out hundreds or thousands of citations within just a few seconds. Trouble is, 99% of them are useless. And that's a generous estimate. We need to have a better system for categorizing the Web so useful stuff can be found -- something more like a library catalog. That won't be easy, though, because on the Web, there is no one in charge of what libraries call accession -- bringing stuff into the collection. Anyone who wants can just add to the Web's resources. That's the Web's strongest point and its weakest point, simultaneously.
BobABW: Try this next, from KAITLY16. (Did your last comment answer this entirely, Peter?)
Question: Why is there so much traffic?
PeterCoyBW: Why is there so much traffic? Because of people like us. Can't tell people not to do stuff on the Internet, can we?
BobABW: And now, from Tapit, who is referring, I think, to faster modems.
Question: Will the new modem speeds help the jamming?
PeterCoyBW: We got into that question a little earlier. The answer is that the faster modem may help you a little bit, but it could also add to the congestion on the Net as a whole.
BobABW: IrvLustig seems to have a point here, Peter. What about the danger to the phone companies from Internet phoning? And what about the traffic jam that could cause?
Question: MCI and Sprint want the slowdown because of Internet phone eroding their business.
PeterCoyBW: Now I see the point you were getting at earlier. And as I said a bit earlier, I think that MCI and Sprint will always be able to make money if they can bring their costs down faster than their prices. And Internet phoning could be cheap for them to deliver. I'm not saying this is an easy thing to do, but I suspect it's the wave of the future.
BobABW: Dependable MikeMerc from South Plainfield, N.J., weighs in next. Didn't you actually suggest this yourself?
Question: Peter, what do you think about proposals that people pay for their Net usage depending on how many bytes they access or transmit? Can that be done? If so, what kind of pay scales might there be?
PeterCoyBW: O.K., that's a very big issue. It would require too much overhead to charge per byte. But there might be a way to come up with a rough equivalent. In fact, that's the concept behind, say, per-minute pricing schemes. They are a very rough mechanism for pricing on the basis of usage. I definitely feel that in the long term, prices must be based on costs -- at least in broad terms. Otherwise, people will find a way to exploit an underpriced resource to the detriment of others.
BobABW: Speaking of Internet fees, profits, things like that, here's Cuckoo again, with a uniquely American comment.
Comment: But there are thousands and thousands of small, flexible local services competing on their reliability and the quality of their equipment. Won't the free market take care of a lot of the problems? ($300 million marketing budgets notwithstanding.)
PeterCoyBW: The small local services can't really control what happens on the backbone of the Net, nor can they control what the content providers do. So, no, they can't solve the problems no matter how good a job they do.
BobABW: Up next, JackB.
Question: When full audio and video are available to all on the Net, will that exacerbate the traffic problems because of the volume of data?
PeterCoyBW: Simple answer: Most definitely yes.
BobABW: Peter, try this, a reaction to your story.
Question: Your poll suggested that people aren't terribly upset yet with their Net experiences. How do you explain that?
BobABW: And tell us what poll that was.
PeterCoyBW: I'm glad someone brought up the BW/Harris Poll [accompanying the Aug. 26 story]. The poll said that most Internet users are satisfied with their Net experiences. To me, that's great news. But what we don't know is whether they are the brave pioneers who don't mind catching a few rocks on the head. I suspect that the broader public -- the ones watching TV at this moment -- would not be as tolerant as we all are.
BobABW: Cuckoo95 raises another interesting point with this comment, Peter. Why don't ISPs just have plenty of backup?
Question: AOL doesn't have "sufficient redundancy," Peter! Otherwise, the system wouldn't have gone offline for 19 hours! They had no backup system!
PeterCoyBW: That comment is obviously on the mark. I wasn't trying to defend AOL's performance in the blackout -- far from it. I was only responding to someone's statement that having all your servers in one place is inherently a bad idea. My answer: That ain't necessarily so.
BobABW: Peter, you have at least partly answer ScottyNY already, but just in case not.
Question: I remember AOL's growing pains too. Don't we just need a magnitude jump in bandwidth to match the jump in users?
PeterCoyBW: If bandwidth were the only problem, that would be fine. But there are other issues -- including the searchability of the Web, which it appears none of our participants want to discuss.
BobABW: Now here's a question it'll take a telecom expert to answer -- from QSFox.
Question: With high-speed modems, won't the I/Net go with bursty transmission like the military?
PeterCoyBW: I need to know more of what you're talking about. Obviously, packet switching is inherently suited for "bursty" transmission.
BobABW: Send a follow-up, QS. Next, a sensitive point.
Question: Considering the loads the Net's being asked to carry, doesn't someone have to step in with the equivalent of the interstate highway program, but who else besides the government could do that?
PeterCoyBW: I feel that is unnecessary. Indeed, the National Science Foundation wisely reduced its role in running the Internet backbone. I would hate to have the Internet caught up in the budget-cutting going on in Washington right now. Can you imagine if it had to fight for its place in a budget against tobacco price supports?
BobABW: Here's ScottyNY's opinion on any proposal for charging on the Net. Peter, agree or not?
Comment: No way on pay scales. Advertisers are more than happy to pay the bills!
PeterCoyBW: Advertisers will pay bills for people logging onto sites where they advertise. But that can't cover all of the Internet's costs.
BobABW: Scotty will make one exception.
Question: Specialized databases can always charge for access.
PeterCoyBW: Fair point.
BobABW: Here's the elaboration from QS Fox.
Question: At 2400 your content rate is slow. At 28.8 the content rate increases. Under the new cable system, won't bursts of 2 or 3 MB be the norm?
PeterCoyBW: Yes, they will. That will put a big strain on the Net.
BobABW: Peter, more observations from Cuckoo.
Comment: Important note on that poll: It surveyed *INTERNET* users, not users of commercial online services (like America Online). A good chunk of the Net's traffic comes from AOL (many of whose problems are unique to AOL).
BobABW: What say you?
PeterCoyBW: Lots of people use terms fairly loosely. So anyone who sends E-mail to someone outside of AOL, or uses AOL to surf the Web, could be considered an Internet user. We weren't able to delve deeply into how people defined their terms, but I'm sure we got a bunch of AOL subscribers in the mix.
BobABW: Back to the free market for a moment, Peter, with this question from PUSHMI.
Question: How expensive is it to increase capacity? Won't providers (AT&T, etc.) have an incentive to do so?
PeterCoyBW: The, or maybe one, big problem, I think, is in local phone companies providing connections to the Internet for people connecting by modem, as most of us are at the moment. See, they have to tie up a whole circuit for each of us. They're used to providing circuits for people to do quick phone calls. But people who go online often stay on for hours at a time. It completely throws off the phone companies' traffic engineering. And don't forget, most of us don't pay for local calls. So they're facing a huge additional cost burden without any additional compensation. That's the kind of thing that needs to change -- will change.
BobABW: And a follow from PUSHMI.
Question: What percentage of Net users are AOL users? Any idea?
PeterCoyBW: Sorry, I'd be guessing.
BobABW: O.K., folks, I think what we'll do is thank Peter for an illuminating discussion on Net traffic jams, present and future.
PeterCoyBW: Did I set a record for the most verbiage spilled in one conference?
BobABW: Maybe not the most, but among the most interesting. Thanks, Peter, and to all of you in the audience for your many questions.
PeterCoyBW: Thank you. Enjoyed it. Goodnight, all.
BobABW: Thanks again, Peter. And goodnight.
Copyright 1996 by The McGraw-Hill Companies, Inc. All rights reserved.
